Kristo Paiba

Built a production Generative AI platform using LLMs and PyTorch, enabling real-time interpretation of industrial IoT data and reducing false-alert noise by 32%. Designed an agentic AI maintenance assistant using LangChain, RAG, and FAISS to give technicians prompt answers. Automated engineering reports with LORA-tuned GPT models and TensorRT, reducing manual reporting effort by 70%. Created a high-speed, multi-agent inference pipeline using FastAPI and ONNX Runtime, improving system latency and throughput. Combined YOLOv8 and OpenCV with LLM reasoning to generate clinical diagnostics from image and sensor inputs. Implemented full MLOps automation with MLflow, Airflow, and Weights & Biases for training, versioning, evaluation, and deployment. Mentored engineers on prompt design, LLM lifecycle management, and best practices for AI-driven product development.

Developed an LLM-based personalization engine using Transformers and scikit-learn, improving recommendation accuracy and user engagement. Built Campaign Content Generator, a multilingual generative system powered by GPT APIs and LangChain. Delivered low-latency inference services with FastAPI and ONNX Runtime at scale. Mentored junior developers in LLM evaluation, prompt engineering, and experiment tracking.

Built backend for a logistics automation platform using Python, Django, and Flask. Developed automated features and REST APIs using Celery and Redis, increasing backend throughput by 40%. Deployed services with Docker and collaborated with cross-functional teams in Agile sprints.